Skip to content

Latest commit

 

History

History
155 lines (122 loc) · 5.2 KB

README.md

File metadata and controls

155 lines (122 loc) · 5.2 KB
Raw

PyNAM

PyNAM is a spiking neural network implementation of the Willshaw associative memory (BiNAM), implemented in Python with PyNNLess/PyNN as a backend.

It serves as a potential benchmark for neuromorphic hardware systems and as a test bed for experiments with the Willshaw associative memory.

Usage

To run the complete analysis pipeline, run the program in one of the following forms:

./run.py <SIMULATOR> [<EXPERIMENT>]
./run.py <SIMULATOR> --process <EXPERIMENT>
./run.py <SIMULATOR> --process-keep <EXPERIMENT>

Where SIMULATOR is the simulator that should be used for execution (see below) and EXPERIMENT is a JSON file describing the experiment that should be executed. If EXPERIMENT is not given, the program will try to execute "experiment.json". The "process-keep" mode will keep all intermediate spiking output. The "process-keep" mode is equivalent to calling ./run with "--create", "--exec" and "--analyse" in a row.

In order to just generate the network descriptions for a specific experiment, use the following format:

./run.py <SIMULATOR> --create <EXPERIMENT>

Note that the SIMULATOR is needed in order to partition the experiment according to the hardware resources. This command will generate a set of ".in.gz" files that can be passed to the execution stage:

./run.py <SIMULATOR> --exec <IN_1> ... <IN_N>

This command will execute the network simulations for the given network descriptors -- if more than one network descriptor is given, a new process will be spawned for each execution. Generates a series of ".out.gz" files in the same directory as the input files.

./run.py <TARGET> --analyse <OUT_1> ... <OUT_N>

Analyses the given output files, generates a HDF5/Matlab file as TARGET containing the processed results.

If the expected output data is expected to be large the following commands can be used:

./run.py <SIMULATOR> --analyse-exec <IN_1> ... <IN_N>
./run.py <TARGET> --analyse-join <OUT_1> ... <OUT_N>

The "analyse-exec" mode will directly generate HDF5/Matlab files after each execution. These files are much smaller than the intermediate files generated by "exec". All results can then be joined using the "analyse-join" mode.

Simulators

Possible simulators are:

  • spikey
  • nest
  • nmmc1
  • nmpm1
  • ess

Experiment description documentation

The experiment description is read from a JSON-like file format -- in contrast to standard JSON, we allow JavaScript-comments inside the file. The available parameters are as follows:

{
    /*
     * The data parameters specify the data that is stored in the Willshaw
     * associative memory and implicitly the network size.
     */
    "data": {
        "n_bits_in": 16, // Number of input bits
        "n_bits_out": 16, // Corresponds to the number of neurons
        "n_ones_in": 3, // Number of ones in the input samples
        "n_ones_out": 3 // Number of ones in the output samples
        //"n_samples": 100 // Number of samples (set automatically if not given)
    },

    /**
     * Network topology and neuron parameters
     */
    "topology": {
        "multiplicity": 1, // Size of a neuron population
        // PyNN neuron parameters (depending on the current model)
        "params": {
            "cm": 0.2,
            "e_rev_E": -40,
            "e_rev_I": -60,
            "v_rest": -50,
            "v_reset": -70,
            "v_thresh": -47
        },
        "param_noise": {
            // Standard deviation of each neuron parameter
        },
        // Neuron model, either IF_cond_exp or EIF_cond_exp_isfa_ista
        "neuron_type": "IF_cond_exp",
        "w": 0.011,  // Synapse weight
        "sigma_w": 0.0 // Standard deviation of the synapse weight
    },

    /**
     * Input data specification
     */
    "input": {
        "burst_size": 1, // Number of spikes in an input burst
        "time_window": 100.0, // Time between samples
        "isi": 1.0, // Time between spikes in a burst
        "sigma_t": 0.0, // Standard deviation of the spike time noise
        "sigma_t_offs": 0.0, // Standard deviation of the burst offset
        "p0": 0.0, // Probability of input spikes missing
        "p1": 0.0 // Probability of a (partial) false-positive input burst
    },

    /**
     * Output data specification
     */
    "output": {
        "burst_size": 1 // Number of spikes representing a "1"
    },

    /**
     * List of experiments to be conducted
     */
    "experiments": [
        {
            "name": "sweep_w", // Name of the experiment
            "sweeps": { // Map of dimensions that should be swept over
                "topology.w": {"min": 0.0, "max": 0.01, "count": 10}
            },
            "repeat": 10 // Number of times the experiment should be repeated
        }
    ]
}

Authors

This project has been initiated by Andreas Stöckel in 2015 as part of his Masters Thesis at Bielefeld University in the Cognitronics and Sensor Systems Group which is part of the Human Brain Project, SP 9.

License

This project and all its files are licensed under the GPL version 3 unless explicitly stated differently.